Total cost of ownership (HD5850 vs HD4890)

[Vertibird]

Is HD4890 really still a viable choice for the new buyer?

1. HD5850 requires the same number of power connectors. (Anyone who can run HD4890 on their computer can run HD5850 also)
2. HD5850 uses far less power at idle and load. (possibly making its total cost of ownership actually lower than HD4890)
3. HD5850 is just plain faster than HD4890.
4. HD5850 is more future proof. (DX11, etc)

If a person wants lower performance wouldn't HD5770 or HD5750 make more sense?

 

[Schmide]

1. Considering with tricks you can get the 4890 into the $150 range, it still meets a price point.
2. At 20-40w difference, it would take 2-3 years of constant usage to make up the cost difference.
3. It's faster and better but by how much.
4. The future is the future, you could invest that $100 in a lemonade stand.

 

[Vertibird]

Originally posted by: Schmide

2. At 20-40w difference, it would take 2-3 years of constant usage to make up the cost .

The difference at idle is more like 50 watts. I think the load power difference is even more.

If the person buying lives in California I'm sure total cost of ownership for HD4890 vs HD5850 would be even higher. (re: greater kw/h electricity costs)

 

[Schmide]

xbitlabs does the best power tests.

California 15c a kwh

Idle 43.7w - 14.7w = 30w = $39.42 a year
2d 53.3w - 25.8w = 27.5w = $36.14 a year
Load 120.3w - 107w = 13.3w = $17.48 a year

See you save money by running it load 24/7 (kidding)

 

[MundoX]

I think 5850 deserves your money. Even if it's 40% more than 4890.

 

[MundoX]

IF 5830 will come, than it would kill all the price competitors. Also 5890. Those should come this year, till then i'll pass.

 

[Patrickz0rs]

I am sitting in the buyers seat right now. I have everything ready to build expect a GPU. This is what I have brought it down to:

1. Get a 4890 if you think DX11 will take a long time OR you don't care about DX11. Also look for deals to clear shelves.

2. Get a 5770 if you want less performance than a 4890 and you're on a budget. The 4890 beats 5770 on DX11 games.

3. Get a 5850 if you want more performance than a 4890 and you have a larger budget. Try to get a deal on one around $220ish?

If I could find a 5850 for under $250 shipped I would be all over it. If a 5830 comes out I will be on that quick too.

 

[RussianSensation]

Considering Sapphire 4890 is $169.99 with free shipping and it easily outperforms 5770 for just $10 more, I don't see what the problem is?

If anything GTX260 216, GTX 275, 5770s and 5850s are all overpriced.

At $259.99, 5850 is 50% more expensive while delivering only 16% more performance on average at 1920x1200 4AA/16AF. 4890 is an amazing deal right now, easily undercutting GTX275 by $30-$35.

Ati needs to lower pricing on 5770 to $130 and 5850 to $219 if anything. But what they will likely do is slowly let 4870 1GB and 4890 1GB fade away and in a couple months release something at $199 between 5770 and 5850. I think in a month or 2 we will see 5850 closer to $239 mark? But who knows.

 

[RussianSensation]

Originally posted by: MundoX
I think 5850 deserves your money. Even if it's 40% more than 4890.

Why? It doesn't perform even 30% faster on average, but right now it's priced at least 50% more.

 

[potato28]

I recently had to make this decision, and I bought the 4890. Why? It fit my price point better and it was easier to get locally, if I'm looking at buying a DX11 card I'll buy the next ones that come out and play the first DX11 games in DX10.

 

[Acanthus]

Im a new 4890 1GB owner.

They are cheaper, and i'm on a tight budget.

Nothing else even came close around $160.

 

[Forumpanda]

You have to factor in total system cost as well, if you are building your system for gaming, if a card cost 50% more and delivers 30% more performance, it is probably still better for a new gaming system as it will only make the system about 10% more expesive but still give 30% more performance.

 

[nicnas]

I was hoping to buy a 5770 but then saw it has 128 bit memory bus resulting in 73.6 GB/s (4890 has 124.8 GB/s). Later I thought I'll wait for the reviews maybe it will be fine.

Today I eagerly read the rewievs and MEH ... it's slower than 4870 (not a big surprise).

4890 is excellent even today, but too noisy and power-hungry for me.

May be I'll hold on to my ancient and slow 2600XT until ATI launch 5830 (256 bit bus please) or hope nVidia get it right with mainstream Fermi ... another 6 months gaming on minimum settings 0xAA 0xAF @1920x1080

So if you have a 1920x1200(1080) monitor (as I do), 600+W power supply with 2 PCI-e connectors (not my case) and don't mind having a noisy graphics card the 4890 is still a great buy IMO. The Sapphire Vapor-X are great (less noise), but for a bit more you can get 5850!

I'm so sad ATI was always the mainstream king but now both ATI and nVidia left the mainstream throne empty (ah the good old 8800GT, 9600GT, 4850, 4770 days).

 

[RussianSensation]

Originally posted by: Forumpanda
You have to factor in total system cost as well, if you are building your system for gaming, if a card cost 50% more and delivers 30% more performance, it is probably still better for a new gaming system as it will only make the system about 10% more expesive but still give 30% more performance.

That's a good point. At the same time you may have sold your 9800GTX for $65 and purchased a 4890 for $170, resulting in a net outlay of $100. If you sold a 9800GTX for $65 and purchased a $260 5850, you paid almost $200 for the upgrade. In this case you pay 100% more for 30% more performance.

I think we are comparing the 2 cards on a standalone basis to each other. If someone asked me if they should purchase a Core i7 860 and 4890 vs. Core i5 750 and 5850, of course I'd recommend the latter.

Relative arguments are relative

 

[BenSkywalker]

Ati needs to lower pricing on 5770 to $130 and 5850 to $219 if anything.

Out of all the new parts ATi has launched, the 5850 is by far the best value, it certainly does not need a price cut at the moment. The 5x70 parts look like they could shed a few bucks, but the 5x50 parts are looking like solid values atm, particularly the 5850.

 

[Vertibird]

Originally posted by: Schmide
xbitlabs does the best power tests.

California 15c a kwh

Idle 43.7w - 14.7w = 30w = $39.42 a year
2d 53.3w - 25.8w = 27.5w = $36.14 a year
Load 120.3w - 107w = 13.3w = $17.48 a year

See you save money by running it load 24/7 (kidding)

Some people in California pay 25c a kwh and Anandtech got 45 watts difference at idle comparing HD5850 to HD4890. In the worst case scenario a HD5850 is a much cheaper product to own.

 

[Vertibird]

Originally posted by: RussianSensation

Ati needs to lower pricing on 5770 to $130 and 5850 to $219 if anything. But what they will likely do is slowly let 4870 1GB and 4890 1GB fade away and in a couple months release something at $199 between 5770 and 5850. I think in a month or 2 we will see 5850 closer to $239 mark? But who knows.

If HD5830 requires two power connectors then comparing it to HD4890 is apples to apples.

The typical buyer for HD5770 has a limited PSU and owns a HD4850 or GTS 250 (at best).

 

[Vertibird]

Originally posted by: Forumpanda
You have to factor in total system cost as well, if you are building your system for gaming, if a card cost 50% more and delivers 30% more performance, it is probably still better for a new gaming system as it will only make the system about 10% more expesive but still give 30% more performance.

Not only that but frame rates increases don't always paint a clear picture of a GPUs power.

Maybe the CPU was bottlenecking the card being tested? HD5850 is a much stronger card than HD4890.

 

[Schmide]

Originally posted by: Vertibird
Some people in California pay 25c a kwh and Anandtech got 45 watts difference at idle comparing HD5850 to HD4890. In the worst case scenario a HD5850 is a much cheaper product to own.

You're arguing a caveat that is contingent on floating variables.

How many people run their gaming rigs 24/7?

I've lived in California, yes you can pay as much as 25c an hour, but that's only in extreme usage situations where you use a large amount on their graduated rate scale. In most cases if you average the amount with the rate you usually never really pay more than 18c a kwh. Take your bill add up the 3 rate entries and divide by the total kwh. :light:

Now it's much cheaper? At the very least it's nearly equal, with a great deal of effort to make it so.

 

[RussianSensation]

Originally posted by: BenSkywalker

Ati needs to lower pricing on 5770 to $130 and 5850 to $219 if anything.

Out of all the new parts ATi has launched, the 5850 is by far the best value, it certainly does not need a price cut at the moment. The 5x70 parts look like they could shed a few bucks, but the 5x50 parts are looking like solid values atm, particularly the 5850.

It may be the best value from the new performance cards, but not the best value overall at the moment.

I think $135 4870 1GB is a great value card. And we know that 2 of these will be faster than single 5870. So 5850 cannot possibly be the best value since its priced even more than dual 4870s. The only reason 5850 is such a good value is because GTX285 is priced completely out of whack.

As has been stated before, 4850 debuted at $199, while 4870 came out at $299. So 5850 should been been priced closer to $199 given its current performance.

But in the future with DX11, it should start to significantly pull away from 4890s, etc. I think when we compare the performance increase of 4850 over 3850 and 4870 over 3870, the performance of both 5850 and 5870 is lacklustre.

 

[cusideabelincoln]

Originally posted by: Schmide
xbitlabs does the best power tests.

California 15c a kwh

Idle 43.7w - 14.7w = 30w = $39.42 a year
2d 53.3w - 25.8w = 27.5w = $36.14 a year
Load 120.3w - 107w = 13.3w = $17.48 a year

See you save money by running it load 24/7 (kidding)

I like Techpowerup's better. Their numbers seem more accurate than Xbitlabs. http://www.techpowerup.com/reviews/HIS/HD_5850/28.html

42W difference at idle.

41-43W difference under 3D gaming situations.

Anandtech also shows similar difference: http://www.anandtech.com/video/showdoc.aspx?i=3650&p=13

46W at idle

45W at gaming load


At this difference, if the computer is turned on 12 hours a day every day for a year, then at $0.18/kwhr a person would save roughly $33.

Obviously if the computer is turned on less, the number would be lower. Consequently, if it's turned on more or if you calculate for a longer period of time, the number would be higher.

 

[BenSkywalker]

It may be the best value from the new performance cards, but not the best value overall at the moment.

It's the best value out of all the new parts, easily.

I think $135 4870 1GB is a great value card. And we know that 2 of these will be faster than single 5870. So 5850 cannot possibly be the best value since its priced even more than dual 4870s.

Where exactly are you shopping? $260 before rebate for the 5850, it isn't more then 2x 4870 1GBs even if you include the rebates(which normally you would only be able to use one rebate per household). Also, using that same logic, 2x9800GTs tend to be faster then a 4890, and cheaper, but given how close they are overall not many people would make the argument as having a single GPU still has some rather clear advantages.

As has been stated before, 4850 debuted at $199, while 4870 came out at $299. So 5850 should been been priced closer to $199 given its current performance.

The 4850 was the fourth fastest(and not much ahead of the 5th) single GPU when it came out. The 5850 is the second fastest by a rather comfortable margin that makes a big difference.

 

[Vertibird]

Originally posted by: cusideabelincoln

Originally posted by: Schmide
xbitlabs does the best power tests.

California 15c a kwh

Idle 43.7w - 14.7w = 30w = $39.42 a year
2d 53.3w - 25.8w = 27.5w = $36.14 a year
Load 120.3w - 107w = 13.3w = $17.48 a year

See you save money by running it load 24/7 (kidding)

I like Techpowerup's better. Their numbers seem more accurate than Xbitlabs. http://www.techpowerup.com/reviews/HIS/HD_5850/28.html

42W difference at idle.

41-43W difference under 3D gaming situations.

Anandtech also shows similar difference: http://www.anandtech.com/video/showdoc.aspx?i=3650&p=13

46W at idle

45W at gaming load


At this difference, if the computer is turned on 12 hours a day every day for a year, then at $0.18/kwhr a person would save roughly $33.

Obviously if the computer is turned on less, the number would be lower. Consequently, if it's turned on more or if you calculate for a longer period of time, the number would be higher.

Even at 12 hours/day this helps put things in perspective.

Now if we compared two Crossfired HD4870 1GB (which is an even worse power hog than 4890 by 10 watts) to a single 5850 it becomes a complete blowout favoring the new technology.

 

[edplayer]

Originally posted by: RussianSensation
Considering Sapphire 4890 is $169.99 with free shipping and it easily outperforms 5770 for just $10 more, I don't see what the problem is?

If anything GTX260 216, GTX 275, 5770s and 5850s are all overpriced.

I can get a BFG GTX260 for $130 right now (AR). Does the 4890 average 30% better? If not, it is the 4890 that is overpriced.

 

[Vertibird]

Originally posted by: RussianSensation

But in the future with DX11, it should start to significantly pull away from 4890s, etc. I think when we compare the performance increase of 4850 over 3850 and 4870 over 3870, the performance of both 5850 and 5870 is lacklustre.

DX11 simplifying graphics computations makes sense.

When CPUs get faster, the difference between 4890 and 5850 will also be more noticeable.

5850 is 50% more TFLOPs than 4890 (both at stock speeds)

5850 is 80% more TFLOPs than 4890 (both at 1000 Mhz core)

Speaking of really powerful GPUs, I wonder what is going to happen to GT300 when it gets compared to 58xx using the same Core i7 processor?

Even though GT300 will be 50% more powerful than HD58xx, I'm sure the gains will look merely incremental in comparison graphs. (Really high resolutions through triple monitors will be a different story though)